Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models

<p dir="ltr">Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial pe...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Hassan Ali (3348749) (author)
مؤلفون آخرون: Muhammad Atif Butt (10849980) (author), Fethi Filali (12646471) (author), Ala Al-Fuqaha (4434340) (author), Junaid Qadir (16494902) (author)
منشور في: 2023
الموضوعات:
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1864513510171475968
author Hassan Ali (3348749)
author2 Muhammad Atif Butt (10849980)
Fethi Filali (12646471)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
author2_role author
author
author
author
author_facet Hassan Ali (3348749)
Muhammad Atif Butt (10849980)
Fethi Filali (12646471)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
author_role author
dc.creator.none.fl_str_mv Hassan Ali (3348749)
Muhammad Atif Butt (10849980)
Fethi Filali (12646471)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
dc.date.none.fl_str_mv 2023-12-28T18:00:00Z
dc.identifier.none.fl_str_mv 10.1109/tits.2023.3343971
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Consistent_Valid_Physically-Realizable_Adversarial_Attack_Against_Crowd-Flow_Prediction_Models/26393263
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Information and computing sciences
Artificial intelligence
Cybersecurity and privacy
Data management and data science
Machine learning
Deep neural networks
CFP
adversarial ML
Perturbation methods
Standards
Adaptation models
Computer architecture
Analytical models
History
Data models
dc.title.none.fl_str_mv Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial perturbations. Although many works have studied these adversarial perturbations in general, the adversarial vulnerabilities of deep CFP models in particular have remained largely unexplored. In this paper, we perform a rigorous analysis of the adversarial vulnerabilities of DL-based CFP models under multiple threat settings, making three-fold contributions; 1) we propose CaV-detect by formally identifying two novel properties— C onsistency a nd V alidity—of the CFP inputs that enable the detect ion of standard adversarial inputs with 0% false acceptance rate (FAR); 2) we leverage universal adversarial perturbations and an adaptive adversarial loss to present adaptive adversarial attacks to evade CaV-detect defense; 3) we propose CVP, a C onsistent, V alid and P hysically-realizable adversarial attack, that explicitly inducts the consistency and validity priors in the perturbation generation mechanism. We find out that although the crowd-flow models are vulnerable to adversarial perturbations, it is extremely challenging to simulate these perturbations in physical settings, notably when CaV-detect is in place. We also show that CVP attack considerably outperforms the adaptively modified standard attacks in FAR and adversarial loss metrics. We conclude with useful insights emerging from our work and highlight promising future research directions.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Transactions on Intelligent Transportation Systems<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/tits.2023.3343971" target="_blank">https://dx.doi.org/10.1109/tits.2023.3343971</a></p>
eu_rights_str_mv openAccess
id Manara2_8578eb17f70a7f6a51101c12cf2063f0
identifier_str_mv 10.1109/tits.2023.3343971
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/26393263
publishDate 2023
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction ModelsHassan Ali (3348749)Muhammad Atif Butt (10849980)Fethi Filali (12646471)Ala Al-Fuqaha (4434340)Junaid Qadir (16494902)Information and computing sciencesArtificial intelligenceCybersecurity and privacyData management and data scienceMachine learningDeep neural networksCFPadversarial MLPerturbation methodsStandardsAdaptation modelsComputer architectureAnalytical modelsHistoryData models<p dir="ltr">Recent works have shown that deep learning (DL) models can effectively learn city-wide crowd-flow patterns, which can be used for more effective urban planning and smart city management. However, DL models have been known to perform poorly on inconspicuous adversarial perturbations. Although many works have studied these adversarial perturbations in general, the adversarial vulnerabilities of deep CFP models in particular have remained largely unexplored. In this paper, we perform a rigorous analysis of the adversarial vulnerabilities of DL-based CFP models under multiple threat settings, making three-fold contributions; 1) we propose CaV-detect by formally identifying two novel properties— C onsistency a nd V alidity—of the CFP inputs that enable the detect ion of standard adversarial inputs with 0% false acceptance rate (FAR); 2) we leverage universal adversarial perturbations and an adaptive adversarial loss to present adaptive adversarial attacks to evade CaV-detect defense; 3) we propose CVP, a C onsistent, V alid and P hysically-realizable adversarial attack, that explicitly inducts the consistency and validity priors in the perturbation generation mechanism. We find out that although the crowd-flow models are vulnerable to adversarial perturbations, it is extremely challenging to simulate these perturbations in physical settings, notably when CaV-detect is in place. We also show that CVP attack considerably outperforms the adaptively modified standard attacks in FAR and adversarial loss metrics. We conclude with useful insights emerging from our work and highlight promising future research directions.</p><h2>Other Information</h2><p dir="ltr">Published in: IEEE Transactions on Intelligent Transportation Systems<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1109/tits.2023.3343971" target="_blank">https://dx.doi.org/10.1109/tits.2023.3343971</a></p>2023-12-28T18:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1109/tits.2023.3343971https://figshare.com/articles/journal_contribution/Consistent_Valid_Physically-Realizable_Adversarial_Attack_Against_Crowd-Flow_Prediction_Models/26393263CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/263932632023-12-28T18:00:00Z
spellingShingle Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
Hassan Ali (3348749)
Information and computing sciences
Artificial intelligence
Cybersecurity and privacy
Data management and data science
Machine learning
Deep neural networks
CFP
adversarial ML
Perturbation methods
Standards
Adaptation models
Computer architecture
Analytical models
History
Data models
status_str publishedVersion
title Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
title_full Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
title_fullStr Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
title_full_unstemmed Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
title_short Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
title_sort Consistent Valid Physically-Realizable Adversarial Attack Against Crowd-Flow Prediction Models
topic Information and computing sciences
Artificial intelligence
Cybersecurity and privacy
Data management and data science
Machine learning
Deep neural networks
CFP
adversarial ML
Perturbation methods
Standards
Adaptation models
Computer architecture
Analytical models
History
Data models